The objective measurement method of minimum resolvable temperature difference for infrared imaging system based on ANFIS

Minimum resolvable temperature difference (MRTD) is an important index for the infrared imaging system. In practice, the measurement of MRTD is observing the images obtained through testing platform by human eyes. This method is unstable and time consuming as it's susceptible to psychological and physiological factors of human. In this paper, an MRTD objective measurement method based on adaptive neuro-fuzzy inference system (ANFIS) is proposed, which is combined with subtractive clustering algorithm to reduce training time. Through analyzing the process of the human eye measurement method, the proposed method regards the MRTD measurement as a classification problem of temperature differences. At first, the infrared images are preprocessed by the average multi-frame images method to simulate human eye visual characteristic and features are extracted to represent the type of images. The features extracted from infrared images at different temperature differences are labeled by the experienced professionals, and the labels divide temperature differences into three categories. Then the features are taken as inputs and the corresponding labels are regarded as target outputs to train ANFIS model for the prediction of MRTD for infrared imaging system. Comprehensive experiments demonstrate this study explore the application of ANFIS to the objective measurement of MRTD for infrared imaging system, and compare the proposed method with backpropagation neural network (BPNN) and human eye method. The experiment results prove the effectiveness and accuracy of the proposed objective MRTD measurement method.